RefereeBench - Entity Perception: leaderboard
RefereeBench (2026) tests multimodal models as sports referees on 925 officiating videos across 11 sports with 6,475 human-annotated multiple-choice questions. This board is accuracy in percent on questions about which players or objects are involved, averaged across sports; higher is better.
Metric: Accuracy (%) on entity-perception questions (Q6), averaged across 11 sports; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 11 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 79.2 | #180 |
| 2 | Gemini 3 Pro | 78.6 | #146 |
| 3 | Seed 1.8 | 76.8 | #258 |
| 4 | GPT-5 | 72.6 | #185 |
| 5 | Qwen 3 VL 8B | 67.4 | #927 |
| 6 | GPT-4o | 65.3 | #754 |
| 7 | InternVL3.5-8B | 61.8 | #1618 |
| 8 | Claude Sonnet 4.5 | 59.1 | #267 |
| 9 | Claude Haiku 4.5 | 55.4 | #587 |
| 10 | VideoLLaMA3-7B | 54.8 | #2064 |
| 11 | LLaVA-Video-7B | 26.7 | #2303 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=refereebench-entity-perception · How It Works · Data refreshed daily, snapshot 2026-11-22.